CODEWITHSUNDEEP
pythonpandas
user2314737
user2314737

Reputation: 29307

Error when computing mode of a DataFrameGroupBy object

I have a dataframe with a Date column, I group the data by year and I can compute mean and median. But how to compute the mode? Here is the error I get:

>>> np.random.seed(0)
>>> rng = pd.date_range('2010-01-01', periods=10, freq='2M')
>>> df = pd.DataFrame({ 'Date': rng, 'Val': np.random.random_integers(0,100,size=10) })
>>> df
        Date  Val
0 2010-01-31   44
1 2010-03-31   47
2 2010-05-31   64
3 2010-07-31   67
4 2010-09-30   67
5 2010-11-30    9
6 2011-01-31   83
7 2011-03-31   21
8 2011-05-31   36
9 2011-07-31   87
>>> df.groupby(pd.Grouper(key='Date',freq='A')).mean()
                  Val
Date                 
2010-12-31  49.666667
2011-12-31  56.750000
>>> df.groupby(pd.Grouper(key='Date',freq='A')).median()
             Val
Date            
2010-12-31  55.5
2011-12-31  59.5
>>> df.groupby(pd.Grouper(key='Date',freq='A')).mode()

Traceback (most recent call last):
  File "<pyshell#109>", line 1, in <module>
    df.groupby(pd.Grouper(key='Date',freq='A')).mode()
  File "C:\Python27\lib\site-packages\pandas\core\groupby.py", line 554, in __getattr__
    return self._make_wrapper(attr)
  File "C:\Python27\lib\site-packages\pandas\core\groupby.py", line 571, in _make_wrapper
    raise AttributeError(msg)
AttributeError: Cannot access callable attribute 'mode' of 'DataFrameGroupBy' objects, try using the 'apply' method

Upvotes: 5

Views: 10902

Answers (3)

Vaid&#248;tas I.
Vaid&#248;tas I.

Reputation: 554

mode is problematic as others have mentioned, however a DataFrameGroupby object can be applied a trivial lambda function, just as the AttributeError suggests using (and contains no ugly slicing or anything else):

df.groupby(grouping_column)[[i for i in pivotable_columns]].apply(lambda x: x.mode())]

Upvotes: 0

piRSquared
piRSquared

Reputation: 294218

  • use np.unique with the return_counts parameter.
  • use the argmax on the counts array to get value from unique array.
  • use np.apply_along_axis for a custom function mode
def mode(a):
    u, c = np.unique(a, return_counts=True)
    return u[c.argmax()]

df.groupby(pd.Grouper(key='Date',freq='A')).Val.apply(mode)

Date
2010-12-31    67
2011-12-31    21
Freq: A-DEC, Name: Val, dtype: int64

Upvotes: 2

3novak
3novak

Reputation: 2544

mode isn't a built in function that's automatically compatible with pandas groupby objects. You could use the scipy.stats module. This feels a little clunky, though.

from scipy import stats

df.groupby(pd.Grouper(key='Date',freq='A')).apply(stats.mode)

Alternatively, you could use the value_counts() function and take the first index value returned. This is the route I would go.

df.groupby(pd.Grouper(key='Date', freq='A')).value_counts()[0].index.values[0]

Upvotes: 1

Related Questions